Merge pull request #1 from declare-lab/humaneval

Humaneval

.gitignore

@@ -1 +1,2 @@
 /.idea/
+*.jsonl

README.md

@@ -65,6 +65,15 @@ python main.py bbh --model_name llama --model_path eachadea/vicuna-13b --load_8b
 # 0.3717117791946168
 ```
 
+Evaluate on [Human Eval](https://github.com/openai/human-eval/tree/master/data) which includes 164 coding questions
+```
+python humaneval.py main --model_name seq_to_seq --model_path google/flan-t5-xl
+
+python humaneval.py main  --model_name llama --model_path decapoda-research/llama-7b-hf
+
+python humaneval.py main  --model_name llama --model_path chavinlo/alpaca-native
+
+```
 ### Setup
 
 Install dependencies and download data.
@@ -76,4 +85,4 @@ pip install -r requirements.txt
 mkdir -p data
 wget https://people.eecs.berkeley.edu/~hendrycks/data.tar -O data/mmlu.tar
 tar -xf data/mmlu.tar -C data && mv data/data data/mmlu
-```
+```

human-eval/LICENSE

@@ -0,0 +1,21 @@
+The MIT License
+
+Copyright (c) OpenAI (https://openai.com)
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

human-eval/README.md

@@ -0,0 +1,115 @@
+# HumanEval: Hand-Written Evaluation Set 
+
+This is an evaluation harness for the HumanEval problem solving dataset
+described in the paper "[Evaluating Large Language Models Trained on
+Code](https://arxiv.org/abs/2107.03374)".
+
+## Installation
+
+Make sure to use python 3.7 or later:
+```
+$ conda create -n codex python=3.7
+$ conda activate codex
+```
+
+Check out and install this repository:
+```
+$ git clone https://github.com/openai/human-eval
+$ pip install -e human-eval
+```
+
+## Usage
+
+**This program exists to run untrusted model-generated code. Users are strongly
+encouraged not to do so outside of a robust security sandbox. The [execution
+call](https://github.com/openai/human-eval/blob/master/human_eval/execution.py#L48-L58)
+in `execution.py` is deliberately commented out to ensure users read this
+disclaimer before running code in a potentially unsafe manner. See the comment in
+`execution.py` for more information and instructions.**
+
+After following the above instructions to enable execution, generate samples
+and save them in the following JSON Lines (jsonl) format, where each sample is
+formatted into a single line like so:
+```
+{"task_id": "Corresponding HumanEval task ID", "completion": "Completion only without the prompt"}
+```
+We provide `example_problem.jsonl` and `example_solutions.jsonl` under `data`
+to illustrate the format and help with debugging.
+
+Here is nearly functional example code (you just have to provide
+`generate_one_completion` to make it work) that saves generated completions to
+`samples.jsonl`.
+```
+from human_eval.data import write_jsonl, read_problems
+
+problems = read_problems()
+
+num_samples_per_task = 200
+samples = [
+    dict(task_id=task_id, completion=generate_one_completion(problems[task_id]["prompt"]))
+    for task_id in problems
+    for _ in range(num_samples_per_task)
+]
+write_jsonl("samples.jsonl", samples)
+```
+
+To evaluate the samples, run
+```
+$ evaluate_functional_correctness samples.jsonl
+Reading samples...
+32800it [00:01, 23787.50it/s]
+Running test suites...
+100%|...| 32800/32800 [16:11<00:00, 33.76it/s]
+Writing results to samples.jsonl_results.jsonl...
+100%|...| 32800/32800 [00:00<00:00, 42876.84it/s]
+{'pass@1': ..., 'pass@10': ..., 'pass@100': ...}
+```
+This script provides more fine-grained information in a new file ending in
+`<input_path>_results.jsonl`. Each row now contains whether the completion
+`passed` along with the execution `result` which is one of "passed", "timed
+out", or "failed".
+
+As a quick sanity-check, the example samples should yield 0.5 pass@1.
+```
+$ evaluate_functional_correctness data/example_samples.jsonl --problem_file=data/example_problem.jsonl
+Reading samples...
+6it [00:00, 3397.11it/s]
+Running example suites...
+100%|...| 6/6 [00:03<00:00,  1.96it/s]
+Writing results to data/example_samples.jsonl_results.jsonl...
+100%|...| 6/6 [00:00<00:00, 6148.50it/s]
+{'pass@1': 0.4999999999999999}
+```
+
+Because there is no unbiased way of estimating pass@k when there are fewer
+samples than k, the script does not evaluate pass@k for these cases. To
+evaluate with other k values, pass `--k=<comma-separated-values-here>`. For
+other options, see
+```
+$ evaluate_functional_correctness --help
+```
+However, we recommend that you use the default values for the rest.
+
+## Known Issues
+
+While evaluation uses very little memory, you might see the following error
+message when the system is running out of RAM. Since this may cause some
+correct programs to fail, we recommend that you free some memory and try again.
+```
+malloc: can't allocate region
+```
+
+## Citation
+
+Please cite using the following bibtex entry:
+
+```
+@article{chen2021codex,
+  title={Evaluating Large Language Models Trained on Code},
+  author={Mark Chen and Jerry Tworek and Heewoo Jun and Qiming Yuan and Henrique Ponde de Oliveira Pinto and Jared Kaplan and Harri Edwards and Yuri Burda and Nicholas Joseph and Greg Brockman and Alex Ray and Raul Puri and Gretchen Krueger and Michael Petrov and Heidy Khlaaf and Girish Sastry and Pamela Mishkin and Brooke Chan and Scott Gray and Nick Ryder and Mikhail Pavlov and Alethea Power and Lukasz Kaiser and Mohammad Bavarian and Clemens Winter and Philippe Tillet and Felipe Petroski Such and Dave Cummings and Matthias Plappert and Fotios Chantzis and Elizabeth Barnes and Ariel Herbert-Voss and William Hebgen Guss and Alex Nichol and Alex Paino and Nikolas Tezak and Jie Tang and Igor Babuschkin and Suchir Balaji and Shantanu Jain and William Saunders and Christopher Hesse and Andrew N. Carr and Jan Leike and Josh Achiam and Vedant Misra and Evan Morikawa and Alec Radford and Matthew Knight and Miles Brundage and Mira Murati and Katie Mayer and Peter Welinder and Bob McGrew and Dario Amodei and Sam McCandlish and Ilya Sutskever and Wojciech Zaremba},
+  year={2021},
+  eprint={2107.03374},
+  archivePrefix={arXiv},
+  primaryClass={cs.LG}
+}
+```

human-eval/data/example_problem.jsonl

@@ -0,0 +1 @@
+{"task_id": "test/0", "prompt": "def return1():\n", "canonical_solution": "    return 1", "test": "def check(candidate):\n    assert candidate() == 1", "entry_point": "return1"}

human-eval/data/example_samples.jsonl

@@ -0,0 +1,6 @@
+{"task_id": "test/0", "completion": "    import subprocess\n    subprocess.check_output('rm -rf tmp')"}
+{"task_id": "test/0", "completion": "    import time\n    time.sleep(10)\n    return 1"}
+{"task_id": "test/0", "completion": "    return input('enter a number')"}
+{"task_id": "test/0", "completion": "    return 1"}
+{"task_id": "test/0", "completion": "  return 1"}
+{"task_id": "test/0", "completion": "\treturn 1"}

human-eval/human_eval/data.py

@@ -0,0 +1,49 @@
+from typing import Iterable, Dict
+import gzip
+import json
+import os
+
+
+ROOT = os.path.dirname(os.path.abspath(__file__))
+HUMAN_EVAL = os.path.join(ROOT, "..", "data", "HumanEval.jsonl.gz")
+
+
+def read_problems(evalset_file: str = HUMAN_EVAL) -> Dict[str, Dict]:
+    return {task["task_id"]: task for task in stream_jsonl(evalset_file)}
+
+
+def stream_jsonl(filename: str) -> Iterable[Dict]:
+    """
+    Parses each jsonl line and yields it as a dictionary
+    """
+    if filename.endswith(".gz"):
+        with open(filename, "rb") as gzfp:
+            with gzip.open(gzfp, 'rt') as fp:
+                for line in fp:
+                    if any(not x.isspace() for x in line):
+                        yield json.loads(line)
+    else:
+        with open(filename, "r") as fp:
+            for line in fp:
+                if any(not x.isspace() for x in line):
+                    yield json.loads(line)
+
+
+def write_jsonl(filename: str, data: Iterable[Dict], append: bool = False):
+    """
+    Writes an iterable of dictionaries to jsonl
+    """
+    if append:
+        mode = 'ab'
+    else:
+        mode = 'wb'
+    filename = os.path.expanduser(filename)
+    if filename.endswith(".gz"):
+        with open(filename, mode) as fp:
+            with gzip.GzipFile(fileobj=fp, mode='wb') as gzfp:
+                for x in data:
+                    gzfp.write((json.dumps(x) + "\n").encode('utf-8'))
+    else:
+        with open(filename, mode) as fp:
+            for x in data:
+                fp.write((json.dumps(x) + "\n").encode('utf-8'))

human-eval/human_eval/evaluate_functional_correctness.py

@@ -0,0 +1,23 @@
+import fire
+import sys
+
+from human_eval.data import HUMAN_EVAL
+from human_eval.evaluation import evaluate_functional_correctness
+
+
+def entry_point(
+    sample_file: str,
+    k: str = "1,10,100",
+    n_workers: int = 4,
+    timeout: float = 3.0,
+    problem_file: str = HUMAN_EVAL,
+):
+    """
+    Evaluates the functional correctness of generated samples, and writes
+    results to f"{sample_file}_results.jsonl.gz"
+    """
+    k = list(map(int, k.split(",")))
+    results = evaluate_functional_correctness(sample_file, k, n_workers, timeout, problem_file)
+
+    return results
+

human-eval/human_eval/evaluation.py

@@ -0,0 +1,105 @@
+from collections import defaultdict, Counter
+from concurrent.futures import ThreadPoolExecutor, as_completed
+from typing import List, Union, Iterable, Dict
+import itertools
+
+import numpy as np
+import tqdm
+
+from human_eval.data import HUMAN_EVAL, read_problems, stream_jsonl, write_jsonl
+from human_eval.execution import check_correctness
+
+
+def estimate_pass_at_k(
+    num_samples: Union[int, List[int], np.ndarray],
+    num_correct: Union[List[int], np.ndarray],
+    k: int
+) -> np.ndarray:
+    """
+    Estimates pass@k of each problem and returns them in an array.
+    """
+
+    def estimator(n: int, c: int, k: int) -> float:
+        """
+        Calculates 1 - comb(n - c, k) / comb(n, k).
+        """
+        if n - c < k:
+            return 1.0
+        return 1.0 - np.prod(1.0 - k / np.arange(n - c + 1, n + 1))
+
+    if isinstance(num_samples, int):
+        num_samples_it = itertools.repeat(num_samples, len(num_correct))
+    else:
+        assert len(num_samples) == len(num_correct)
+        num_samples_it = iter(num_samples)
+
+    return np.array([estimator(int(n), int(c), k) for n, c in zip(num_samples_it, num_correct)])
+
+
+def evaluate_functional_correctness(
+    sample_file: str,
+    k: List[int] = [1, 10, 100],
+    n_workers: int = 4,
+    timeout: float = 3.0,
+    problem_file: str = HUMAN_EVAL,
+):
+    """
+    Evaluates the functional correctness of generated samples, and writes
+    results to f"{sample_file}_results.jsonl.gz"
+    """
+
+    problems = read_problems(problem_file)
+
+    # Check the generated samples against test suites.
+    with ThreadPoolExecutor(max_workers=n_workers) as executor:
+
+        futures = []
+        completion_id = Counter()
+        n_samples = 0
+        results = defaultdict(list)
+
+        print("Reading samples...")
+        for sample in tqdm.tqdm(stream_jsonl(sample_file)):
+            task_id = sample["task_id"]
+            completion = sample["completion"]
+            args = (problems[task_id], completion, timeout, completion_id[task_id])
+            future = executor.submit(check_correctness, *args)
+            futures.append(future)
+            completion_id[task_id] += 1
+            n_samples += 1
+
+        assert len(completion_id) == len(problems), "Some problems are not attempted."
+
+        print("Running test suites...")
+        for future in tqdm.tqdm(as_completed(futures), total=len(futures)):
+            result = future.result()
+            results[result["task_id"]].append((result["completion_id"], result))
+
+    # Calculate pass@k.
+    total, correct = [], []
+    for result in results.values():
+        result.sort()
+        passed = [r[1]["passed"] for r in result]
+        total.append(len(passed))
+        correct.append(sum(passed))
+    total = np.array(total)
+    correct = np.array(correct)
+
+    ks = k
+    pass_at_k = {f"pass@{k}": estimate_pass_at_k(total, correct, k).mean()
+                 for k in ks if (total >= k).all()}
+
+    # Finally, save the results in one file:
+    def combine_results():
+        for sample in stream_jsonl(sample_file):
+            task_id = sample["task_id"]
+            result = results[task_id].pop(0)
+            sample["result"] = result[1]["result"]
+            sample["passed"] = result[1]["passed"]
+            yield sample
+
+    out_file = sample_file + "_results.jsonl"
+    print(f"Writing results to {out_file}...")
+    write_jsonl(out_file, tqdm.tqdm(combine_results(), total=n_samples))
+
+    return pass_at_k